Optimizing Mode Connectivity for Class Incremental Learning

Haitao Wen, Haoyang Cheng, Heqian Qiu, Lanxiao Wang, Lili Pan, Hongliang Li
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:36940-36957, 2023.

Abstract

Class incremental learning (CIL) is one of the most challenging scenarios in continual learning. Existing work mainly focuses on strategies like memory replay, regularization, or dynamic architecture but ignores a crucial aspect: mode connectivity. Recent studies have shown that different minima can be connected by a low-loss valley, and ensembling over the valley shows improved performance and robustness. Motivated by this, we try to investigate the connectivity in CIL and find that the high-loss ridge exists along the linear connection between two adjacent continual minima. To dodge the ridge, we propose parameter-saving OPtimizing Connectivity (OPC) based on Fourier series and gradient projection for finding the low-loss path between minima. The optimized path provides infinite low-loss solutions. We further propose EOPC to ensemble points within a local bent cylinder to improve performance on learned tasks. Our scheme can serve as a plug-in unit, extensive experiments on CIFAR-100, ImageNet-100, and ImageNet-1K show consistent improvements when adapting EOPC to existing representative CIL methods. Our code is available at https://github.com/HaitaoWen/EOPC.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-wen23b, title = {Optimizing Mode Connectivity for Class Incremental Learning}, author = {Wen, Haitao and Cheng, Haoyang and Qiu, Heqian and Wang, Lanxiao and Pan, Lili and Li, Hongliang}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {36940--36957}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/wen23b/wen23b.pdf}, url = {https://proceedings.mlr.press/v202/wen23b.html}, abstract = {Class incremental learning (CIL) is one of the most challenging scenarios in continual learning. Existing work mainly focuses on strategies like memory replay, regularization, or dynamic architecture but ignores a crucial aspect: mode connectivity. Recent studies have shown that different minima can be connected by a low-loss valley, and ensembling over the valley shows improved performance and robustness. Motivated by this, we try to investigate the connectivity in CIL and find that the high-loss ridge exists along the linear connection between two adjacent continual minima. To dodge the ridge, we propose parameter-saving OPtimizing Connectivity (OPC) based on Fourier series and gradient projection for finding the low-loss path between minima. The optimized path provides infinite low-loss solutions. We further propose EOPC to ensemble points within a local bent cylinder to improve performance on learned tasks. Our scheme can serve as a plug-in unit, extensive experiments on CIFAR-100, ImageNet-100, and ImageNet-1K show consistent improvements when adapting EOPC to existing representative CIL methods. Our code is available at https://github.com/HaitaoWen/EOPC.} }
Endnote
%0 Conference Paper %T Optimizing Mode Connectivity for Class Incremental Learning %A Haitao Wen %A Haoyang Cheng %A Heqian Qiu %A Lanxiao Wang %A Lili Pan %A Hongliang Li %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-wen23b %I PMLR %P 36940--36957 %U https://proceedings.mlr.press/v202/wen23b.html %V 202 %X Class incremental learning (CIL) is one of the most challenging scenarios in continual learning. Existing work mainly focuses on strategies like memory replay, regularization, or dynamic architecture but ignores a crucial aspect: mode connectivity. Recent studies have shown that different minima can be connected by a low-loss valley, and ensembling over the valley shows improved performance and robustness. Motivated by this, we try to investigate the connectivity in CIL and find that the high-loss ridge exists along the linear connection between two adjacent continual minima. To dodge the ridge, we propose parameter-saving OPtimizing Connectivity (OPC) based on Fourier series and gradient projection for finding the low-loss path between minima. The optimized path provides infinite low-loss solutions. We further propose EOPC to ensemble points within a local bent cylinder to improve performance on learned tasks. Our scheme can serve as a plug-in unit, extensive experiments on CIFAR-100, ImageNet-100, and ImageNet-1K show consistent improvements when adapting EOPC to existing representative CIL methods. Our code is available at https://github.com/HaitaoWen/EOPC.
APA
Wen, H., Cheng, H., Qiu, H., Wang, L., Pan, L. & Li, H.. (2023). Optimizing Mode Connectivity for Class Incremental Learning. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:36940-36957 Available from https://proceedings.mlr.press/v202/wen23b.html.

Related Material